A low memory bandwidth Gaussian mixture model (GMM) processor for 20, 000-word real-time speech recognition FPGA system.
Kazuo MiuraHiroki NoguchiHiroshi KawaguchiMasahiko YoshimotoPublished in: FPT (2008)
Keyphrases
- gaussian mixture model
- speech recognition
- low memory
- speaker identification
- speech recognizer
- speech recognition systems
- speaker recognition
- mixture model
- language model
- em algorithm
- speech signal
- hidden markov models
- feature vectors
- expectation maximization
- gaussian mixture modeling
- n gram
- background subtraction
- automatic speech recognition
- maximum likelihood
- pattern recognition
- noisy environments
- memory requirements
- probabilistic neural network
- signal processing
- feature space
- computer vision
- mel frequency cepstral coefficients
- gaussian distribution
- word recognition
- word segmentation
- frequency domain